- Published on
Canada's AI strategy shouldn't include secret Palantir bills
- Authors
- Name
- DeepHorus
Canada's newly released "AI for All" strategy, which positions the federal government as a "strategic anchor customer" for domestic artificial intelligence, faces criticism over its lack of transparency and continued reliance on foreign vendors. Policy analyst Al Vigier argues that while the strategy promises to boost Canadian AI adoption from 12 percent to 60 percent by 2034, the government has been quietly purchasing American systems, notably from Palantir Technologies, without public disclosure.
The Department of National Defence signed a contract with Palantir's Canadian arm in March 2020, initially valued at $14.4 million but never publicly announced. Through more than a dozen amendments, the contract's worth climbed to approximately $44.4 million by October 2024, with actual spending reaching $46.8 million. A separate $3.7 million Defence contract only surfaced after a Conservative MP questioned AI spending. Additionally, the Ontario Provincial Police have used Palantir's Gotham platform since 2015.
These contracts involve data-fusion and decision-support systems that the new strategy claims Canada must develop domestically. Vigier notes that the strategy allocates $500 million for equity stakes in promising AI firms, $700 million for compute power, a Trusted AI Certification program, and a health-focused missions program. However, none of these provisions include direct purchase orders from Canadian vendors.
The equity approach may deter private investment, as government stakes can create uncertainty for allied investors. Vigier warns that state-built national champions often become wards of the state rather than competitive market players. Furthermore, starting the missions program in healthcareāa sector known for slow procurement and high privacy concernsārisks stalling progress.
Critics argue that the strategy should prioritize transparent procurement from Canadian small and medium enterprises, rather than relying on equity or complex certification schemes. Vigier suggests making source lists accessible to companies with fewer than 50 employees and setting a minimum floor for domestic procurement.
The true test of the anchor customer model will be whether the government issues open, competitive contracts to Canadian firms. Without a clear mandate to buy domestically, the strategy may fail to build the sovereign AI base it promises, leaving Canada dependent on foreign systems acquired behind closed doors.
